This study derived data from the Investigations on interpersonal adversity, psychiatric status, and socio-cognitive function (IAPS) project, which aimed to investigate the interactions between interpersonal adversity, psychiatric status, level of dissociation, and socio-cognitive functions [13]. The IAPS project recruited healthy controls and inpatients with MDD, BD with depressive episodes, and schizophrenia. The recruited inpatients were assessed with two phases of examinations. Phase 1 assessments were conducted at the first week of admission. Self-reported measures of psychopathology and childhood interpersonal adversity, neurocognitive test, and clinical interview schedules were recorded. In addition, phase 2 assessments were the same as in phase 1, and they were conducted during the last week of admission or 6 weeks after the admission. Assessment of healthy controls were conducted only in Phase 1. This research was approved by the Institutional Review Board of Kai-Syuan Psychiatric Hospital (KSPH-2014-34 and KSPH-2017-04) and conducted according to the current revision of the Declaration of Helsinki and national legal requirements (Human Subjects Research Act, Taiwan). All participants signed written informed consent after a detailed explanation of the study procedures.

Details of the IAPS project have been published elsewhere [13,23,24]. In this study, we included data of newly admitted patients with the acute onset of MDD and BD at Kaohsiung Municipal Kai-Syuan Psychiatric Hospital, Taiwan, from December 30, 2014 to December 21, 2016. The inclusion criteria of patients were: 1) admission due to MDD or BD following the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV-TR or DSM-5 diagnostic criteria through interviews by psychiatrists; 2) age between 20 and 50 years; and 3) a native speaker of Mandarin Chinese. Patients were not eligible for this study if they: 1) had been diagnosed as ketamine use disorder, or 2) exhibited intellectual disability so that they could not understand the purpose of the IAPS project or follow orders of the researchers to complete the tests. Intellectual disability was diagnosed from boarded psychiatrists and was recorded in the medical record. We excluded them in the initial screen. The recruited patients were assessed with two series of examinations. The baseline assessments were conducted during the first week of admission, and the follow-up assessments were conducted during the last week of admission or 6 weeks after admission. During the admission, the patients were treated with standardized treatment for MDD or BD by psychiatrists.

Descriptive statistics were applied to summarize the clinical characteristics at baseline. The patients were divided into MDD and BD groups, and the two groups were compared. Pearson’s χ² test was used to compare categorical variables, and the independent t-test was used for continuous variables. To estimate the differences in the PANSS-P, CADSS, and HAMD, we used the independent t-test to compare scores between baseline and follow-up. Moreover, we analyzed differences in the changes in PANSS-P, CADSS, and HAMD scores between the high-dissociation group and the low-dissociation group. Since no previous study has defined the cut-off point of the CADSS to differentiate high and low levels of dissociation, we used the mean score (14.82) as the cut-off point. The independent t-test was used to compare baseline and follow-up scores between the high- and low-dissociation groups.

To estimate the factors associated with changes in CADSS scores over time, generalized estimating equations (GEEs) with a first-order autoregressive working correlation structure [33] were used; these GEEs were also used to estimate the interrelation between depression and dissociation. If changes in HAMD scores were associated with changes in CADSS scores over time, GEEs were used again to identify the predictive effect of each item of the HAMD (17 items) on the change in CADSS scores. To reduce the confounding effect, each item of the HAMD in the analysis was adjusted for multiple characteristics, including sex, comorbidities (PTSD, BPD, and substance abuse), age, level of education, IQ scores, DDD (antidepressants, antipsychotics, and benzodiazepine), and severity of psychotic symptoms. All tests were two-tailed, and statistical significance was set at p<0.05. All data were processed using SPSS version 23.0 for Windows (IBM Corp.).

This study comprehensively explores the association between dissociation and depression across symptomology. We found differences in the changes of the severity of depression, psychotic symptoms, and dissociation over time, and the severity of depression and psychotic symptoms significantly decreased at follow-up. However, although the level of dissociation did not significantly decrease in the overall cohort or lowdissociation group, it did significantly decrease in the highdissociation group. In addition, we identified that the change in the level of dissociation was positively associated with the change in the level of depression. Moreover, the change in psychotic symptoms demonstrated a similar pattern to the level of depression. Females and those who were comorbid with PTSD or BPD were also associated with an increased level of dissociation. After multiple adjustments, more severe levels of middle insomnia and gastrointestinal symptoms were associated with an increased level of dissociation. Early insomnia and anxiety somatic symptoms demonstrated similar but insignificant trends. However, a decreased level of late insomnia was associated with an increased level of dissociation, indicating that those who had a lower level of late insomnia may have had a higher level of dissociation.

We found some controversial but interesting findings for the association between depression and dissociation. The level of depression significantly decreased after standardized treatment, but the level of dissociation did not. This may indicate that dissociation is etiologically different from depression, since standardized treatment for depression was not as effective for dissociation. Some studies have also demonstrated that comorbidity with dissociation may interfere with the treatment efficacy of mental illnesses such as BPD or anxiety [6,20]. This is consistent with our findings that dissociation may be a unique psychopathology that compromises the treatment efficacy of other mental illnesses. Intriguingly, GEE analysis showed that the change in depression was positively correlated with the change in dissociation, and that this correlation may be due to the associated symptoms of depression. We found that the significant predictors of dissociation were insomnia and gastrointestinal symptoms, but not depressed mood. Taken together, we hypothesize that the association between depression and dissociation may be due to insomnia or gastrointestinal symptoms, but not the core symptom of MDD (depressed mood) [34]. This may also reflect an etiological difference between dissociation and depression. Nevertheless, further studies are needed to better clarify the complicated etiologies behind dissociation.

Sleep disturbance, including early and middle insomnia, was positively associated with dissociation. Several studies have discussed the impact of sleep disturbance on dissociation, and poor sleep quality and sleepiness have been reported to increase the odds of having dissociative experiences [35]. Another study demonstrated that sleep experiences could explain most variance in the severity of dissociation compared with other factors such as impulsivity or negative affect [36]. Moreover, other research found that unusual dreaming and sleepiness were related to dissociative experiences [37,38]. The finding that late insomnia was negatively correlated with dissociation is unexpected. One previous study reported a negative association between time spent awake during the night and level of dissociation [39], which is comparable to our results. On the other hand, we also found a positive association between dissociation and gastrointestinal upset, including items of gastrointestinal symptoms and anxiety somatic symptoms with subscales of gastrointestinal complaints. A previous meta-analysis indicated a strong link between dissociation and somatic symptom disorder [40]. However, only one study has specifically discussed gastrointestinal symptoms. The authors demonstrated the impact of dissociation on irritable bowel syndrome, and concluded that it may be a long-term coping pattern related to past sexual trauma, emotional distress, alcoholism, and physical disability [40]. Further studies focusing on the association between dissociation and late insomnia or gastrointestinal symptoms may be helpful to clarify the phenomenological characteristics.

We found a positive correlation between dissociation and psychotic symptoms. A previous epidemiological study in the general population demonstrated that dissociation was related to psychotic experiences [8], and another study also reported that a state of dissociation could significantly predict the later occurrence of auditory hallucinations [41]. Our previous work also exhibited this relationship among patients with MDD and BD [13]. In addition, we found that comorbidity with BPD was associated with dissociation. Several studies have comprehensively discussed this association, and it may be related to the socio-cognitive model [42,43]. The socio-cognitive model is an open, dynamic, multidimensional perspective that can accommodate a wide range of social, cultural, and cognitive variables [44]. The socio-cognitive model is thought to contribute to the way that individuals with dissociation express their experiences, such as emotional instability, identity problems, and impulsive behavior, which is associated with BPD [45]. On the other hand, PTSD has been positively correlated with dissociation, which may be explained by the traumatic model of dissociation. According to this model, using dissociation as a defensive strategy to cope with unbearable, overwhelming experiences during a potentially traumatizing event can allegedly compartmentalize experiences and memories in a discrete personality for patients with dissociative disorder [5]. Regarding the impact of gender differences on the change in dissociation, we found that the female participants were positively associated with dissociation. Epidemiological research on children with a history of psychological trauma reported that female children showed a higher level of dissociation symptoms than males [46], and this may echo our results.

There are several limitations to this study. First, the IAPS project did not recruit elderly patients to minimize the confounding effect of recall bias from self-reported questionnaires which were used in other studies. Second, this is a naturalistic observation study, and the recruited participants were treated by various teams. Therefore, the treatment plan of the participants was individualized according to the clinical status and the judgment of their clinical teams rather than standardized. However, we adjusted for the confounding effect of psychotropic medications through their DDD, and this may have minimized this effect.

In conclusion, our results demonstrated a decoupled but intertwined relationship between dissociation and depression. Dissociation should not be considered an epiphenomena of depression, while their symptomology is partially correlated. This correlation may result from the impact of sleep disturbance and gastrointestinal symptoms. In addition, female sex, psychotic symptoms, and comorbidity of BPD were independent factors associated with changes in dissociation. Due to the unique psychopathology of dissociation, the clinical implication of this study is the important and urgent need to develop specific treatment for dissociation, as psychotherapy and psychopharmacological treatment for dissociative disorder are still insufficient [5]. Empirically trauma-specific psychotherapy or some psychotropic medications (e.g., opioid antagonists; paroxetine) may be potentially beneficial in reducing dissociative symptoms [47]. Effective interventions for dissociation may help clinicians to adjust treatment strategies for depression when approaching patients with this comorbidity. Regarding the potential research targets, the link between dissociation and late insomnia or gastrointestinal symptoms among patients with depression deserves further investigation on the clinical symptomology or biomedical etiologies. Moreover, it may extend to other clinical or non-clinical individuals to explore if these factors (sleep problems or gastrointestinal symptoms) are independent from depression. Further studies with well-controlled comparisons and a larger prospective cohort are necessary to explore the precise link and even causality between dissociation and depression.